Research On High-precision Slow-target Localization Algorithms From Angle-only Measurements By Single-platform

Single-platform angle-only target localization has the characteristics of flexible deployment,good concealment,electromagnetic interference resistance and long detection range,and it has been widely used in passive radar target detection,infrared target perception and search and rescue using an autonomous underwater vehicle.With regarding to the applications of long-range target localization,the height information of the observer is unknown,the location of the observer is uncertain,and the measurement noise variance is related to the distance between the target and the observer.How to achieve precision positioning of the slow-target from angle-only measurements in consideration of the above uncertainties is a very challenging topic.In this thesis,a THREE-Stage Extended Kalman Filter(3S-EKF)and a 3S-EKF based particle filter are proposed respectively to deal with the above mentioned uncertainties.The specific research contents are summarized as follows:1.To analyze the model mismatch problem by considering the slow-target as a stationary target,the theoretical bias is derived,followed by demonstrating the relationship between the bias and the velocity of the target.Then a Bayesian Cram(?)r-Rao lower bound(BCRLB)for single-platform angle-only target localization is given.2.A TWO-Stage Extended Kalman Filter(2S-EKF)is proposed to solve the coupling issue between azimuth and elevation angle measurement equations..The proposed2S-EKF algorithm has higher localization performance than the One-Stage Extended Kalman Filter algorithm.Then a new 3S-EKF algorithm is proposed to predict the relative height accurately by employing the triangular relationship between the 2D-projected target position and the relative height.Compared with 2S-EKF algorithm.the 3S-EKF algorithm improves the estimation accuracy of relative height.3.A 3S-EKF based particle filter algorithm is proposed to improve the positioning accuracy when the location of the observer suffers from errors and the variance of measurement noise depends on the distance between observer and target.First,the observer’s position error is mapped into the measurement noise.Then a new proposed distribution formed from the results of the 3S-EKF algorithm is used for particle filtering.Finally,a Bayesian Cram(?)r-Rao lower bound is derived.Compared with the 3S-EKF algorithm,the 3S-EKF based particle filter algorithm provides much higher localization accuracy.